Redesigning Screening with Criteria-Based Intelligence
How I designed a unified AI screening system for IIIP that replaced gut-feel resume scanning with structured, explainable candidate evaluation.
Background - Intervue
Intervue started as an interview outsourcing platform, like Uber for tech interviews. Companies with engineer bandwidth issues would send requests, and we'd assign anonymous external interviewers. But the problem wasn't just interviews. It was the entire hiring infrastructure.
The Origin
Companies were stitching together 6–8 tools just to hire one person, ATS, job boards, assessment platforms, video tools, calendars, spreadsheets. Fragmented data, inconsistent evaluation, slow cycles.
The Evolution
Intervue evolved into IIIP, an all-in-one hiring platform covering AI sourcing, AI screening, interviews, analytics, and offers. I designed every screen across this entire ecosystem.
The Positioning
Where traditional ATS tools make companies switch between platforms, IIIP consolidates the entire pipeline. One job, one platform, one source of truth, from sourcing to offer letter.
The Scale
Currently in beta with 4 enterprise customers using the platform for various hiring needs, from high-volume BPO hiring to specialized tech roles.
Today's Hiring Stack
IIIP Unified Platform
Want the full picture of IIIP?
See how the entire platform fits together across sourcing, screening, interviews, analytics, and offers in the complete IIIP product deck.
The Hiring Funnel is Broken at the Top
Massive candidate volume enters the funnel, but without structured screening, most drop-offs happen at the wrong stages, wasting interviewer time and losing good candidates.
How Recruiters Actually Screen Today
Through conversations with hiring teams and observing recruiter workflows, I identified four core pain points in traditional screening.
No Shared Rubric
Two recruiters looking at the same resume make opposite decisions. There's no standardized criteria evaluation is entirely subjective and varies per person.
Keyword Blindness
ATS systems rely on keyword filtering. A candidate with "distributed systems" expertise gets rejected because they didn't literally write "microservices" on their resume.
No Soft Skill Evaluation
Communication, intent, and cultural fit only surface during HR calls which are expensive and time-consuming. There's no way to assess these signals before the interview stage.
Screening Fatigue
Recruiters batch-process 200+ resumes in a sitting. Quality degrades as the stack grows candidates at the bottom of the pile get less attention than those at the top.
Screening was a filter, not an evaluation. It removed people — but didn’t actually assess them.
The Core Insight
Criteria-Based Screening: The Design Concept
Instead of keyword matching or gut-feel scanning, I designed a system where recruiters define evaluation criteria in plain English, and AI scores candidates against each dimension across resume and video.
- Type criteria the way you think, "3+ years of React"
- No rigid dropdowns or boolean logic
- AI maps the words to evaluation dimensions
- Scored across multiple inputs at once
- Checks the resume and the video intro
- Documents + human signals = one composite score
- Most Fit (80–100), Borderline (40–79), Least Fit (0–39)
- Set your own auto-reject threshold
- Every score is explainable, backed by evidence
- Powers resume, video, and AI sourcing
- Consistent across the whole top-of-funnel
- One set of criteria, three surfaces
Introduction of Screening in the pipeline
A walkthrough of the key interfaces I designed, from criteria configuration to candidate evaluation to pipeline integration.
Same System, Different Surface
The criteria modal adapts for AI sourcing, same plain-English input, but with filters which will be set on backend . This shows the design system's flexibility: one criteria framework serving effective screening during the Applied Phase.
Transparent, Explainable Scoring
Every candidate gets a detailed screening report. "Requirements Matched" shows green checks. "Requirements Not Matched" shows red X marks. The recruiter sees exactly which criteria the resume satisfied and which it didn't. No black-box AI, every score is justified, every decision has evidence.
The Recruiter's Working Dashboard
A filterable table view where recruiters can quickly scan all candidates by fit status. "Most fit" candidates surface first with green tags, "Least fit" with red. Each row links to the detailed screening report, one click from list to evidence.
Extending Criteria to Video
Screening is configured up front, from the Settings tab while creating a job. Recruiters set candidate screening questions, profile fields, and notifications alongside the role's basic details, then layer in evaluation criteria. The same framework spans both Resume screening and AI Video screening tabs: add a video prompt, set a time limit, define plain-English criteria, and tune the auto-reject threshold, two evaluation layers, one unified system, all set before the job goes live.
The Candidate's Side of Video Screening
All the video screening configuration is set on the recruiter side, but this is what the candidate sees after applying. They're nudged to complete a short AI assessment, with applicants who submit video answers surfaced first to companies. That gentle push is what gives recruiters the signal they can't read from a resume: how a candidate actually answers, their tonality, modulation, and energy on camera.
Design Outcomes
Impact measured through design system adoption, workflow efficiency, and structural improvements to the hiring pipeline.
What I Learned
Plain English > Rigid Forms
Giving recruiters free-text criteria input instead of dropdown menus dramatically increased adoption. They could express nuanced requirements ("has worked at a Series B or later startup") that no predefined field could capture.
Transparency Builds Trust in AI
Showing exactly which criteria matched and which didn't, with green checks and red X marks, was the single most important design decision. Recruiters don't trust black-box AI. They trust AI they can verify.
One System, Multiple Surfaces
Designing the criteria framework as a reusable system (not a one-off feature) meant it could power screening, sourcing, and video evaluation simultaneously. This reduced development time and created consistency across the platform.
Empowerment Over Replacement
Recruiters don't want to be replaced by AI, they want to be empowered by it. The "Borderline" tier with manual review was critical. It kept the human in the loop for edge cases while automating the obvious decisions.